Digital Xray Mammography

Digital Xray Mammography THURSDAY MEETING 2.11.2000 Xavi Espinal Curull Need for Digital Mammography • Women has 1/8 chance of developing breast cancer. • Best hope: early detection through mammography. • Detect tumours 2 years before a lump can be felt ( recover chance nearly 100 %). • 20% tumours actually missed in conventional mammographys, even more with young women. • DM advantages: store and exchange pictures, play with contrast, less radiologists fails. • 90 % dose reduction. There is a correlation between several exposures and future cancer development. Xray Project •Low dose Xray Mammography system •CdZnTe Detector •Medipix II Chip Deep sub-micron CMOS process: •Improved spatial resolution than Medipix : pixel size : 55um2 •Preamplifier, comparator and 13 bit counter. •Non-sensitive area as small as posible in order to obtain large areas by tilling the detectors. •Medipix I (PCC): • 32000 counts before threshold • 170 um2 Comparision CCD PCC •Top Image taken by a standard Xray source and a screen + CCD system. ( Indirect Capture ) •Bottom Image, taken by 109Cd source and the PCC. (Direct Capture) •Larger Pixel size in PCC •Inserted needle distinguished in both. •Density differences much more clear in PCC 30 TIMES LOWER DOSE IN PCC IMAGE Absorption Coefficients • CdZnTe has wide range of good absorption for low energies. Cd(1-x)Zn(x)Te Detector • 4000 electron – hole pair per 20KeV  • Good behaviour at room temperature Cd(1-x)Zn(x)Te Detector Growth • • • • Growing High Pressure Bridgman impurities < 1016 cm -3 Te Rich 10-150 bar internal pressure Argon CdZnTe Properties Grown Crystal CdZnTe Properties • High Density: 5.8 g/cm3 and high atomic number. •Provides an excellent absorption coefficient even for thin detectors ( 98% absorption at 20 Kev for 0.4mm thickness ). • High resistivity ( 10-100 Gigaohm per cm ) •Low Dark Currents • High gain 400 e-h pair per 20KeV photon •Excellent signal to noise ratio •Design limitation : Small hole mobility •The Detector must be as thin as possible and biased correctly to provide the shortest distance for the holes to travel. Imaging Process Xray production • Xray Production • Beam Collimator • Interaction with Breast • Energy deposition in the detector • Readout system Colimator BREAST Air GAP DETECTOR Chip Array and Readout electronics Bump Bonding Processes at Breast Level XRay 80% Absorption 10% TrespassingWithout energy Loss 10% Scattered Particles Escaping the Breast Fully absorbed Without Scattering 2,3 or 4 Scatterings But fully absorbed in the detector Arriving with Full energy at the detector The tungsten plane •Tungsten plane is inserted in order to focus the beam in region of 150 um. •Tungsten absorbs all the radiation with a thickness of 0,1mm. •The tungsten Plane / Hole reduces the scattering effects ,which we will discuss later, improving the shape of the image. •Less dose received by the patient as we are focusing only into small regions. Tungsten plane imaging Scattered particles high angle Xray Source Non Scattered particles, and low Angle scattered particles Hole Tungsten Breast Detector •The idea is to take the image in 1 second by making a shift of the hole through all the breast. •So the exposition to the radiation will be low, and the dose will decrease, and this is less invasive for the breast tissue. Contribution to the imaging •Non scattered particles that outcome from the breast with no scatter. •Scattered particles ( roughly speaking ) with low scattering angle. •Scattered particles at CdZnTe detector level. Microcalcifications I • What we are focusing on is to recognise the calcium microcalcifications inside the breast, of about 150um in the early stage. • These are the most common sources of cancerous tumours in women, this microcalcifications are produced by the milk lobuls, in the sacs. Microcalcifications II Calcium Transmission ( 20 KeV ) Absorption ( 20 KeV) 300 um 0.55 0.45 150 um 0.74 0.26 Photon counting • With the chip Medipix2 we are able to count photons in each pixel with an excellent precission. • Simulation results gives the ratio : 0.71 between events beneath Calcium box 150um width, and free space under the breast. • So we have aprox 29% less events under the Calcium phantom. One of the first digital images used to detect breast cancer: Detecting Microcalcifications • A Simulation of (150m)3 calcium box has been made. • The Calcium at 150 um depth stops nearly 26 % of the radiation, theoretically. Detecting Microcalcifications • So there is a difference in the event density between the region beneath calcium phantom, and the same region placed in free space: ratio = 0.7 •This means that if we can control the scattering effects and they are not bigger than 10 % we will be able to detect early microcalcifications. •Moreover we can cut non desired scattering events with energy less than 19 KeV by the resolution of the detector, assuming that we will lose 16% of valid events . ( 18KeV 2.5% lose. ) •Ca Microcalcifications are aprox. Spherical in shape, and the most common are: lobular or low grade ductal carcinoma. Fluka simulations I •Fluctuating kaskad is a Monte Carlo simulation Software developed at CERN by Alfredo Ferrari et al. •This software allows to define several regions to count the energy deposited: breast, detector or whatever at the same run i.e. For each event. •This allows us to track a single photon, and get the number of scatterings that this photon has had, and to know the energy loss in each of the scatterings. •One can make a fine binning of the regions, in order to know with high precission where the photons interact. •We are working with bins of 50um because this is the size of the chip, and there is no need to go to higher binning. •A big trick has been made because for each event we had 8MBytes, so now we are able to generate as many events as we want with no disk quota matters. Fluka simulations I •Several geometries can be implemented, and all the materials and compounds are allowed. Simple Algorithm to detect the image of an Object at the statistical limit of Xray photons flux •Naked eye is limited by nature in detecting digital radiology image, because the statistical noise is dominant. •It Is posible to obtain image with low dose by using a simple algorithm: Sliding-Window Pixel Integration Method •Detection implies a contrast between the N detected beneath the object, and the N elsewhere. •N that reachs the detector underneath the object: (1-C) N •N seen elsewhere: A0N •By computing the difference, we get: A0NC. •At the end of the day , and adding the efficience, , and the error: N   2 (2  C ) A0C 2 • and this formula gives the number of photons needed to make the imaging of an object in order to achieve a certain resolution specified through  and given by multiples of standard desviation: . Imaging low contrast objects Signal to Contrast Ratio: SCR= | n - n’| n Incident Photons Signal to Noise Ratio: SNR= | n - n’| n Object Contrast: ev.dtcted C  1 ev.genrtd n n’ Detector Low contrast detection means : Earlier tumor visibility Patient dose reduction Study of scattering effects in mammography •We are able to determine scattering effects in the process of radiological imaging. •Breast produces scattering at low and high angles. •Simulations have been made in order to quantify the ratio of scattered photons that produce a bad shape of the image. •A box of 150m has been inserted inside the breast tissue , this box was filled with W and Ca. •W at 150 m has a transmission of 0,and Ca with the same width has a transmission of aprox. 0.70. •So, by filling the box with W, we should have no events under the box. But unfortunately we have events. •W absorbs almost all the radiation outcoming from the source. As we have seen in transparence 12. Study of scattering effects in mammography II •Some Results: • There is a dependence on the distance between the object and the detector, the longer the distance, the greater the scattering. Beam W Breast W Shadow •If no scattering we will get no events in the shadow. Study of scattering effects in mammography III •Simulation inputs: •107events generated. •3cm between W and detector. •1cm radiation covered •Box Material: W Events with no W box expected Events obtained Contrast Obtained 700 76 0.89 •We can’t see anything with contrast less than 11% •Box Material: Ca (150um width) Events with no Ca box expected Events obtained Contrast Obtained 700 417 0.41 Low contrast objects Aluminum Thickness Contrast 125 m 9% 100 m 7.2% 75 m 5.4% 40 m 2.9% 25 m 1.8% Photon Counting Detector Film 8 bit 170 m 10.9 mm 255 0 Future Semiconductor Detectors •Mercuric Iodide Detector ( HgI2) •Better performance over a broader energy range than CdZnTe. Material Si Ge CdZnTe HgI2 Z 14 32 48-30-52 Density (g/cm3) 2.33 5.32 5.76 6.30 80-53 •Resistivity is 1000 times bigger than CdZnTe. •Even less dark currents than CdZnTe Conclusions •Xray mammography with 90 % dose reduction. • Able to detect microcalcifications at early stages. • Image can be software optimized. • Reducing the scattering we can improve the image quality: low contrast objects can be imaged.